If a pigeon were to commit good code, I would accept it
- Maxime Garcia (@maxulysse), Jan. 2026
Short summary
nf-core’s policy remains unchanged. Humans are still ultimately responsible for their submitted code.
If you’re using AI tools, try to stick by these guidelines:
- Keep PRs as small and focussed as possible
- Avoid any unnecessary changes, such as moving or refactoring code (unless that is the explicit intention of the PR)
- Review all generated code yourself before opening a PR, and ensure that you understand it
- Engage with the community review process and expect to make revisions
Background
As everyone reading this is most likely aware, Large Language Models (LLMs) have exploded in usage in many areas of society usually under the term artificial intelligence (AI). Particularly in software development, many tools and services have been developed and adopted by many companies, institutes, and developers.
However, the use of AI/LLMs remain a contentious topic, and many conflicting opinions on their use, ethics, and implications remain.
Within nf-core, two discussions have recently taken place regarding the use of AI in code contributions 1 2, with two main viewpoints emerging of strong enthusiasm vs scepticism.
Despite differing opinions and experiences, we reached a consensus on an approach that aims to be as inclusive as possible on how AI and LLMs will be treated within nf-core.
Pros and Cons
To briefly summarise the main arguments for and against the use of AI in code contributions to nf-core.
For the people who were very much in favour, they felt that:
- Their use of AI related tools made their development work more efficient, speeding up development
- The AI tools took over more boring leg work allowing them to work on more fun tasks
- In some cases the use of LLM tools can help with ensuring more consistency in code and quality across the community
- AIs are just ultimately tools, no different to linters, code formatters, or other automated tools already being widely used - it’s up to the human behind the keyboard not to abuse them
On the flip side people who were more sceptical said they worried about:
- AI code and automated reviews having variable and inconsistent quality
- Increased risk of ‘drive-by PRs’ with limited engagement by author
- Increased risk of low-effort but large contributions with significant code changes outside original scope
- Moral, ethical, environmental, and legal questions such as source data attribution remaining unaddressed
- Ultimately risking greater waste of reviewer and community maintenance time
nf-core’s stance
nf-core core team will not ban or prohibit the use of AI or LLM tools in code contributions, nor force any community member to use AI tools.
Our position is that the human using the AI tools remains responsible for any code they submit3.
As with any other code, nf-core community members reserve the right to reject contributions that do not meet our guidelines. This therefore includes cases where PRs, reviews, or other content have been obviously AI-generated without due care. Accordingly, repeated drive-by or low quality PRs will see the same consequences as with any other violation of our contribution guidelines.
We will continue to encourage community members to be open and honest, for example, acknowledging very significant use of AI tools in PR descriptions.
The community will be mindful when assisting members who wish to apply AI tools to nf-core development, that we do not impact the experience of others.
For example, we will aim to not clutter the template with many AI helper files where unnecessary (e.g. AGENT.md, CLAUDE.md, .mcp.json etc.), but will consult with the community via blog posts and RFCs on nf-core/proposals.
We will also aim to further develop and promote best-practice guidelines and etiquette. For example, we hope to find approaches that encourage smaller, focused, and incremental feature PRs - something that will also benefit ‘human’ contributors, such as through faster review response time.
Finally, we will continue to monitor the situation and community feedback. If at any point you have concerns about a PR or possible adoption of AI features in the community, feel free to contact the core team.
Conclusion
To conclude, nothing changes within nf-core. Humans are still responsible for their code, regardless of the tools they use.
This statement was drafted by James Fellows Yates with help from GitHub Co-Pilot Pro’s GPT-4.1 autocompletion. It was reviewed by the nf-core core team. However the image is pure, handmade, artisanal (terrible) ‘photoshopping’.
